Advanced Topics in R

Information Visualization

Rafael S. de Souza

4/17/2017

A statistical computing environment

R integrates data manipulation, graphics and extensive statistical analysis. Uniform documentation and coding standards. But quality control is limited.
Easy download from http://www.r-project.org for Windows, Mac or linux. On-the-fly installation of CRAN packages.
\(>\) 10500 user-provided add-on CRAN packages, tens of thousands of statistical functions
Principal difficulty: Finding what you want, and understanding what you find. Google helps the former problem. Improved education in statistics addresses the latter problem.

Some functionalities of R

Bayesian Inference	Machine Learning	Social Sciences
Computational Physics	Medical Image Analysis	Spatial Data
Cluster Analysis	Multivariate Statistics	Statistical Geneticss
Differential Equations	Natural Language	Survival Analysi
Econometrics	Numerical Mathematic	Time Series Analysis
Environmetrics	Optimization	Visualization
Environmetrics	Pharmacokinetic	Web Technologies
Extreme Value Analysis	Phylogenetics
Empirical Finance	Probability Distributions
Functional Data Analysis	Psychometric

Required packages

require(ggplot2);
require(reshape2);require(d3heatmap);require(circlize);require(ggdendro)

Basic commands

1+1

## [1] 2

x <- 2
for (i in 1:5){
print(x+i)  
}

## [1] 3
## [1] 4
## [1] 5
## [1] 6
## [1] 7

Basic commands

x <- rnorm(100)
hist(x)

Simple regression model

set.seed(1056)                    # set seed to replicate example
nobs= 150                         # number of obs in model 
x1 <- runif(nobs,0,5)             # random uniform variable
mu <- 1 + 5 * x1 - 0.75 * x1 ^ 2  # linear predictor, xb
y <- rnorm(nobs, mu, sd=0.5)      # create y as adjusted random normal variate 
fit <- lm(y ~ x1+I(x1^2))          # Normal Fit

summary(fit)

Fitting linear model: y ~ x1 + I(x1^2)
	Estimate	Std. Error	t value	Pr(>\|t\|)
x1	4.867	0.1057	46.05	3.086e-89
I(x1^2)	-0.7204	0.02074	-34.74	9.438e-73
(Intercept)	1.118	0.1111	10.06	1.92e-18

Plot results

xx <- seq(0,5,length=200)
ypred <- predict(fit,newdata=list(x1=xx),type="response")     # Prediction from the model 

plot(x1,y,pch=19,col="red")                                   # Plot regression line 
lines(xx,ypred,col='cyan',lwd=4,lty=2)

segments(x1,fitted(fit),x1,y,lwd=2,col="gray")                # add the residuals

Data Exploration

Read and display data in table format

d <- read.csv("exoplanets.csv",header = T)
d <- d[complete.cases(d),]
head(d)

Scatter plot

Boxplot

Linear Fit

Histograms

Heatmap

Dendrograms

Chord Diagram

nc <- cor(d[,c(2,3,5,6,7)])
chordDiagram(nc)